Quality of data

In this section I will give an assessment of the quality of my empirical research and design, by explaining how I have tried to achieve reliability, internal and external validity. These three notions can be considered as quality tests, frequently used in social research as this one (Yin 2014). Limitations of the data collected and ethical considerations are presented at last.

3.6.1 Ensuring reliability

Reliability refers to the question if another researcher that applies the same methods as in this study will arrive at the same results (Thagaard 2009). It will be difficult to replicate the study because of its qualitative characteristics, and considering it is an anonymized case and informants. However, as a researcher one can argue for

reliability by explaining how the data has been developed to convince the reader about the quality and value of the results (ibid.). I have therefore throughout the methodology chapter tried to make the process as transparent as possible. Further in this section, I elaborate on accuracy of collected data and the different relations that occurred between the informants and me (Johannesen, Tufte and Christoffersen 2010;

Thagaard 2009). For instance, established trust or not, and potential consequences this may have had for my empirical findings (Thagaard 2009).

To get an accurate as possible collection of data; the interviews were as previously mentioned recorded on a Dictaphone and later transcribed right after each interview to retain a right as possible image of how the interview went. Using a Dictaphone

provided for less inconsistencies and interruption in the information flow compared to taking notes. It was seen as a flexible way to proceed and contributing to a mutual dialog between the informants and me. I felt like I then could fully concentrate on the informants’ answers and come with follow-up questions when necessary. In one of the interviews with the management there were some technical difficulties with the Dictaphone, which was not noticed before after. This caused a data loss of

approximately 10 minutes from the interview, so questions 23-25 in the interview-guide missed. However, I felt I got the essence from the interview despite of this, and from the two other management interviews.

All of the conducted interviews may have been affected by the interview-effect (Johannessen, Tufte and Christoffersen 2010). This effect is related to how I was perceived by the informants. For instance, my attitude, expressions and cloths - all having a potential effect on the answers given by the informants (ibid.). I tried to avoid this effect by having a dress code according to the situation. In interviews with the operative personnel I dressed more casually and informal, and in interviews with the management I had a more formal outfit. In situations when appropriate, I tried to ask for experiences or instances in the past related to themes developing. Such questions often engaged and motivated the informants further because the subjects seemed personal to them and I showed interest. This was part of trying to build an atmosphere of trust, since trust and interview-effect are closely related (Ryen 2002), enchanging the reliability of data collected (Johannessen, Tufte and Christoffersen 2010).

In addition, most of the management interviews were conducted in a natural context, i.e. their offices, which may have made the informants feel more safe and open (Johannesen, Tufte and Christoffersen 2010). With the operative personnel, my contact person appointed an office. This was still within the same organization, but may have created some further distance compared to normally working in the workshops. I dealt with the situation by not sitting directly next to the informant and gave them distance. Humour and drinking coffee was also common to create a more informal setting and seemed to work.

There is no guarantee that the informants always answered what they thought, either because how they wanted to present themselves or how they wanted to be understood (Thagaard 2009). It is reasonable to believe that reporting may have been a sensitive topic for some of the informants, which was noticed in at least one interview based on the non-verbal communication observed. However, since the interviews were face to face I was able to interpret their body language and facial expressions and adapt to the situation. For instance with aforementioned informant, I could read the situation and come back to potentially sensitive questions when it felt more appropriate later on.

This approach provided for more adequate answers and enhanced the quality of data collected (Ringdal 2009).

One and a half hour interview is probably not sufficient time to establish total trust and openness, but the promise for anonymity in each interview may have

compensated for some of this. In addition, particularly the operative informants seemed to open up more as the interview proceeded giving me longer answers

combined with personal experiences, which can be an indicator of some level of trust that I managed to establish.

For me as a researcher I also had to be conscious about asking leading questions during the interviews. There is always a possibility, but I tried my best to avoid this having relatively open follow-up questions in my own interview-guide, as well as sometimes asking improvised questions based on their stories. I regularly sought confirmation through all interviews when longer answers were given, to make sure I understood their answers correctly or to get a more adequate answer (Thagaard 2009).

If there were misunderstandings related to my questions, I asked them in different ways until properly understood by the informants (Ringdal 2009).

Another factor that may have enhanced the reliability of the data is the language. All interviews were conducted in my native language Norwegian, making it easier for me to interpret slang and phrases. On the contrary, the transcribed interviews were later translated into English and some sentences may have lost their original significance.

However, I tried to be aware of this throughout the writing process and citations are used according to the original context of topic.

3.6.2 Ensuring internal validity

Validity is about if I have measured what I actually wanted to measure (Ringdal 2009). In qualitative research as this one, it includes to what degree my selected approaches and findings reflect the study’s purpose and represent the reality (Johannessen, Tufte and Christoffersen 2010). It requires an evaluation of the

interpretations I have made throughout the research project. My position and relations to the case of study also has an affect on how these interpretations are developed (Thagaard 2009).

It has to be made clear that half a year earlier this research project was initiated I had a summer internship in the same organization. However, I feel this has not affected my objectivity since I have had a conscious attitude to it in the whole process of writing the thesis. Most of the interviews were also conducted with informants I did not have any relation to from before. It has rather contributed to possibly better understand the informants’ answers and the conditions of the safety information system since I have had my own experiences in the same organization (Thagaard 2009).

The informants represented all of the different levels within the facility, from operative personnel in the workshops to line-management and management.

Therefore, the informants have provided broader statements and explanations to analyse safety culture and its influence on the SIS. I let the informants speak freely to avoid manipulations and misrepresentations of the case being studied. In addition, several of the informants confirmed each other’s statements throughout the study, indicating that they were speaking sincerely. However, as previously mentioned there is no guarantee that the informants were not withdrawing some information.

3.6.3 Ensuring external validity

External validity means if my research results can be transferred (generalized) to similar phenomena (Johannessen, Tufte and Christoffersen 2010). Case studies are usually criticized for not being able to generalize the empirical findings to larger populations. Generalization is common in quantitative research trying to extrapolate statistical generalizations based upon the empirical findings using larger “sampling units” (Yin 2014). The empirical findings in this single-case study are rather difficult to transfer to similar phenomena, since its characterised by time- and

context-dependency, personal experience and values (Johannessen, Tufte and Christoffersen 2010).

However, it is possible that parts of the study can contribute to a more general theoretical understanding (Thagaard 2009) of safety culture. In chapter 1.2, the study of Oltedal and McArthur (2011) found safety culture to both facilitate and inhibit the reporting frequency within merchant shipping. The authors also applied just

(informed) culture theory of James Reason (1997) to explain how organizations’

handling of blame and trust affects the reporting propensity. Both my and their findings suggest that just culture influences the reporting propensity. More specifically, this study reveals that operative personnel have a low threshold for delivering reports, since they know the management are not going after them and there is a mutual established trust. Thus, it is possibly to transfer some of the findings that have been confirmed through theory to other industries emphasizing safety culture.

3.6.4 Ethical Considerations

There were a number of ethical issues related to this project that had to be taken into consideration. Ethics is about rules, principles and guidelines - a continuous

assessment of actions being right or wrong (Johannessen, Tufte and Christoffersen 2010). The matter of confidentiality is one of these basic ethical principles.

Confidentiality has been sustained by treating the informant’s answers confidential on a password-protected computer until processed and afterwards deleted.

The informants’ privacy and identities have been protected as far as possible (Ryen 2002). There are two ways to identify interview objects: direct identification through

for instance name, or indirect identification through different combinations of information as work title, sex or age (Johannessen, Tufte and Christoffersen 2010).

Both of them were not used in the data collection nor further data reduction, analysis and discussion. All informants were anonymized while writing this thesis and in the text (Thagaard 2009). This was also explained explicitly prior to all conducted interviews with the informants and that my findings were restricted to research purposes only. In addition, specific names and locations that can be linked to this organization have been removed from interview data when transcribed.

Thoughts have been given if the operative personnel could in the aftermath recognize their colleagues that had been interviewed, but this is considered to be less likely. This is because the facility itself is a large one, and the operative personnel volunteered often were from different workshop-locations. There are reasons to believe that the management can easier be recognised internally by colleagues if they try to, due to their close cooperation and fields of expertise.

Chapter 4. Empirical findings